Re: [PATCH] convert uses of ZONE_HIGHMEM to is_highmem

[Andy Whitcroft]

--On 25 June 2004 15:27 -0700 Andrew Morton <akpm@xxxxxxxxx> wrote:

> Andy Whitcroft <apw@xxxxxxxxxxxx> wrote:
> > As the comments in mmzone.h indicate is_highmem() is designed to
> > reduce the proliferation of the constant ZONE_HIGHMEM.  This patch
> > updates three references to ZONE_HIGHMEM to use is_highmem().
> > None appear to be on critical paths.
> >
> > Revision: $Rev: 305 $
> >
> >  void __init set_highmem_pages_init(int bad_ppro)
> >  {
> >  # ifdef CONFIG_HIGHMEM
> > -	int nid;
> > +	struct zone *zone;
> >
> > -	for (nid = 0; nid < numnodes; nid++) {
> > +	for_each_zone(zone) {
> >  		unsigned long node_pfn, node_high_size, zone_start_pfn;
> >  		struct page * zone_mem_map;
> >  		
> > -		node_high_size =
> > NODE_DATA(nid)->node_zones[ZONE_HIGHMEM].spanned_pages; -		zone_mem_map
> > = NODE_DATA(nid)->node_zones[ZONE_HIGHMEM].zone_mem_map; -
> > 	zone_start_pfn =
> > NODE_DATA(nid)->node_zones[ZONE_HIGHMEM].zone_start_pfn; +		if
> > (!is_highmem(zone))
> > +			continue;
> > +
> > +		printk("Initializing %s for node %d\n", zone->name,
> > +			zone->zone_pgdat->node_id);
> > +
> > +		node_high_size = zone->spanned_pages;
> > +		zone_mem_map = zone->zone_mem_map;
> > +		zone_start_pfn = zone->zone_start_pfn;
> >
> > -		printk("Initializing highpages for node %d\n", nid);
> >  		for (node_pfn = 0; node_pfn < node_high_size; node_pfn++) {
> >  			one_highpage_init((struct page *)(zone_mem_map + node_pfn),
> >  					  zone_start_pfn + node_pfn, bad_ppro);
>
> Fair enough.
>
> > @@ -930,11 +930,12 @@ unsigned int nr_free_pagecache_pages(voi
> >  # ifdef CONFIG_HIGHMEM
> >  unsigned int nr_free_highpages (void)
> >  {
> > -	pg_data_t *pgdat;
> > +	struct zone *zone;
> >  	unsigned int pages = 0;
> >
> > -	for_each_pgdat(pgdat)
> > -		pages += pgdat->node_zones[ZONE_HIGHMEM].free_pages;
> > +	for_each_zone(zone)
> > +		if (is_highmem(zone))
> > +			pages += zone->free_pages;
>
> but that's slower.

Yes.  Although I didn't think any of the users were particularly 
performance critical.  The routine counts the number of free pages in the 
zones which count as highmem, ie memory not direct mapped into KVA.  In 
theory there could be more than one zone per node which was highmem, 
ZONE_HIGHMEM is currently the only one.

I'll have a think if there is a more performant way to 'fix' that.

-apw



-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/